9th Annual Conference of the International Speech Communication Association

Brisbane, Australia
September 22-26, 2008

Extraction and Tracking of Formant Response Jitter in the Cochlea for Objective Prediction of SB/SF DAM Attributes

Wenliang Lu, D. Sen

University of New South Wales, Australia

In this paper, we focus on the objective prediction of two of the foreground perceptual quality elements of the Diagnostic Acceptability Measure (DAM) - that of SB and SF - and show that they are correlated with statistical characteristics of features extracted from a physiologically motivated cochlear model response. The work complements earlier work where two other DAM quality elements, SH and SL, were predicted using the same cochlear model [1]. Novel methods of extracting salient features from the cochlear response as well as tracking their evolution are described. Finally, it is shown that the standard deviation of the features is highly correlated with the perception of 'fluttering' (SF) and 'babble' (SB) like distortions.


  1. D. Sen, "Predicting foreground SH, SL and BNH DAM scores for multidimensionalobjective measure of speech quality," Acoustics, Speech, and Signal Processing, 2004. Proceedings. (ICASSP 04). IEEE International Conference on, vol. 1, pp. I-493-6 vol.1, 17-21 May 2004.

Full Paper

Bibliographic reference.  Lu, Wenliang / Sen, D. (2008): "Extraction and tracking of formant response jitter in the cochlea for objective prediction of SB/SF DAM attributes", In INTERSPEECH-2008, 1048-1051.